Overlay and P2P Networks. Introduction and unstructured networks. Prof. Sasu Tarkoma

Transcription

1 Overlay and P2P Networks Introduction and unstructured networks Prof. Sasu Tarkoma

2 Contents Overlay networks and intro to networking Unstructured networks

3 Overlay Networks An overlay network is a network that is built on top of an existing network. The overlay therefore relies on the so called underlay network for basic networking functions, namely routing and forwarding. Today, most overlay networks are built in the application layer on top of the TCP/IP networking suite.

4 Overlay Networks Overlay technologies can be used to overcome some of the limitations of the underlay, and at the same time offering new routing and forwarding features without changing the routers. The nodes in an overlay network are connected via logical links that can span many physical links. A link between two overlay nodes may take several hops in the underlying network.

5 Requirements for Overlay Networks 1. Support the execution of one or more distributed applications by providing infrastructure for them. 2. Participate and support high-level routing and forwarding tasks. The overlay is expected to provide data forwarding capabilities that are different from those that are part of the basic Internet. 3. Deployed across the Internet in such a way that third parties can participate in the organization and operation of the overlay network.

6 Protocol Stack Layers are part of a network architecture Provide services for layers above Hiding the complexity of the current layer Multiple layers are needed in order to reduce complexity Separation of network functions Distribution of complexity OSI, TCP/IP Protocols are building blocks of a network design Can exist independently of layering

7 Background What is network architecture? Layered architecture Network structure Network evolution Motivation for overlay networks

8 Network architecture A set of principles and basic mechanisms that guide network engineering Physical links Communication protocols Format of messages The way in messages are exchanged Elements of the protocol stack Where is the state?

9 Naming, Addressing, and Routing NAMING unicast: to a specific node broadcast: to all nodes multicast: to a subset of nodes anycast: to any one in some subset (IPv6) How to identify and name a node? Even if its address changes. ADDRESSING Where is the node located? ROUTING How to route information to the node s address?

10 TCP/IP Network Stack Application Layer Transport Layer (TCP/UDP) Networking Layer (IP) Underlying network (link layer, physical)

11 Structure of the Internet End-to-End Principle: application logic and state at the edge, core network is simple Network maintains routing tables and forwards packets based on IP prefixes Various protocols for intra-domain routing (OSPF, ) Hierarchical structure: Autonomous Systems with tiers and peering links, economics of inter-domain routing Border Gateway Protocol (BGP) for routing between domains New requirements for IP and network: mobility, multihoming, security, multipath operation,

12 The Hourglass diverse applications Overlays! divergence transport layer (TCP/IP) convergence diverse physical layers

13 Questions Q: Where is the routing state in overlay networks? A: Must be at the edge of the network. Easy to bootstrap, because network is not modified. Q: What is the benefit of this from the network point of view? A: Can overcome limitations of the current network and introduce new services. Q: What is limitation from the network point of view? A: Must take network structure into account (economics). Traversal solutions can be complex.

14 Evolution of the Network Cisco reports trends and estimates in Cisco VNI: Forecast and Methodology, Annual global IP traffic will pass the zettabyte (10 21 ) threshold by the end of 2016 The number of devices connected to IP networks will be nearly three times as high as the global population in Video delivery has become one of the recent services on the Web Estimates of P2P share of network traffic range from 20% to 40% (used to be 40-70%)

15 Cisco forecast: Global Video Highlights Globally, Internet video traffic will be 55 percent of all consumer Internet traffic in 2016, up from 51 percent in The sum of all forms of video (TV, video on demand, Internet, and P2P) will continue to be approximately 86 percent of global consumer traffic by Internet video to TV doubled in Internet video to TV will be 12 percent of consumer Internet video traffic in Video-on-demand traffic will triple by The amount of VoD traffic in 2016 will be equivalent to 4 billion DVDs per month. High-definition video-on-demand surpassed standarddefinition VoD by the end of By 2016, highdefinition Internet video will comprise 79 percent of VoD.

16 With video growth, Internet traffic is evolving from a relatively steady stream of traffic (characteristic of P2P) to a more dynamic traffic pattern.

17 Dynamic Patterns Busy-hour traffic is growing more rapidly than average traffic. Busy-hour traffic will increase nearly fivefold by 2016, while average traffic will increase nearly fourfold. Busy-hour Internet traffic will reach 720 Tbps in 2016, the equivalent of 600 million people streaming a highdefinition video continuously.

18 Network Growth The traffic is expected to increase some 29% each year ( ) Much of this increase comes from the delivery of video data P2P traffic has become a smaller component of Internet traffic in terms of its current share Video is being delivered by a set of protocols, typically coordinated by overlay solutions and CDN solutions We will cover these in during the course

19 CDNs Content Delivery Networks (CDNs) are examples of overlay networks that cache and store content and allow efficient and less costly way to distribute data in massive scale CDNs typically do not require changes to end-systems and they are not peer-to-peer solutions from the viewpoint of the end clients

20 Challenges for Overlay Networks The Real World. In practice, the typical underlay protocol, IP, does not provide universal end-to-end connectivity due to the ubiquitous nature of firewalls and Network Address Translation (NAT) devices. Management and administration. Practical deployment requires that the overlay network has a management interface. Overhead. An overlay network typically consists of a heterogeneous body of devices across the Internet. It is clear that the overlay network cannot be as efficient as the dedicated routers in processing packets and messages. Moreover, the overlay network may not have adequate information about the Internet topology to properly optimize routing processes

21 Layering Routing algorithms Dynamics of content demand and supply Overlay Networks Dynamics of nodes (churn) Network protocols Dynamics of the network, mobility, inter-domain routing

22 Overlay Nodes The overlay nodes are expected to support the network in the following ways: 1. Providing infrastructure for the overlay network and supporting the execution of dis- tributed applications. 2. Participating in the overlay network and perform routing and forwarding of messages and packets. 3. Deployed over the Internet in such a way that they are reachable by third parties.

23 Example Overlay: RON Consider a native link failure in CE Only overlay link AE is affected. The native path AE is rerouted over F (ACE ACFDE) OVERLAY A E 3 G 2 I NATIVE G B A F I Cost Overlay recovery: 8 C + E D H Original: 2 Overlay rerouting: 4 Native Rerouting: 2 Native Failure Native Recovery Time Native Repair

24 Example Overlay: Session Initiation Protocol (SIP) Location Server 4. Where is Bob? 5. Answer: at test.com Proxy 2. Upload location Registrar at test.com 3. SIP Request 6. SIP request 1. REGISTER sip:

25 Network Invariants and Metrics The correctness and performance of a routing algorithm can be analyzed using a number of metrics Typically it is expected that a routing algorithm satisfies certain invariant properties that must be satisfied at all times. The two key properties are safety and liveness The former states that undesired effects do not occur, in other words the algorithm works correctly, and the latter states that the algorithm continues to work correctly, for example avoids deadlocks and loops These properties can typically be proven for a given routing algorithm under certain assumptions Important metrics: shortest path, routing table size, path stretch, forwarding load, churn

26 Trend Challenges Solutions P2P Growth in traffic, upstream bottlenecks P2P caching, localityawareness Internet Broadcast Flash crowds P2P content distribution, multicast technologies Internet Video-on-Demand Growth in traffic, especially metropolitan area and core Content Delivery Networks (CDNs), increasing network capacity, compression Commercial Video-on-Demand Growth in traffic in the metropolitan area network CDNs, increasing network capacity, compression High-definition content Access network IPTV bottleneck, growth in VoD traffic volume in the metropolitan area network CDNs, increasing network capacity, compression

27 Overlay Applications Typical applications of overlay networks include: Content search and file transfer. Distributed directories with efficient lookups. Content routing over the Internet including voice and video. Publish/subscribe and notification. Distributed storage systems. Multi-player games.

28 Terminology Peer-to-peer (P2P) Different from client-server model Each peer has both client/server features Overlay networks Routing systems that run on top of another network, such as the Internet. Distributed Hash Tables (DHT) An algorithm for creating efficient distributed hash tables (lookup structures) Used to implement overlay networks Typical features of P2P / overlays Scalability, resilience, high availability, and they tolerate frequent peer connections and disconnections

29 Peer-to-peer in more detail A P2P system is distributed No centralized control Nodes are symmetric in functionality Large faction of nodes are unreliable Nodes come and go P2P enabled by evolution in data communications and technology Current challenges: Security (zombie networks, trojans), IPR issues P2P systems are decentralized overlays

30 Characteristics of P2P systems P2P can be seen as an organizational principle Applied in many different application domains Characteristics Self-organization Lack of central coordination Resource sharing Based on collaboration between peers Peers are typically equal Large number of peers Resilient to certain kinds of attacks (but vulnerable to others)

31 P2P Volume Estimates range from 20-40% of Internet Traffic Latest estimates from Cisco suggest that video delivery is the growing and the share of P2P traffic is becoming smaller P2P can be used for video delivery as well.. And voice (Skype, P2PSIP) Hundreds of millions of people use P2P technology today

32 Evolution of P2P systems ARPAnet had P2P like qualities End-to-end communication, FTP, USENET,.. Today s BGP is P2P Started from centralized servers Napster Centralized directory Single point of failure Second generation used flooding (Gnutella v0.4) Local directory for each peer High cost, worst-case O(N) messages for lookup Third generation use some structure (Gnutella v0.7) Research systems use DHTs Chord, Tapestry, CAN,.. Decentralization, scalability Some recent CDNs and content delivery systems exhibit P2P features (P2P assisted CDN)